JP5284433B2 - Process monitoring / diagnosis / support equipment - Google Patents

Description

  Embodiments of the present invention assist plant operators in non-stationary situations such as abnormalities in monitoring process systems such as sewage treatment processes, wastewater treatment processes, sludge digestion processes, water purification processes, water supply and distribution processes, chemical processes, and steel processes The present invention relates to a process monitoring / diagnosis / support device that can be used.

  In process plants such as sewage treatment processes, sludge digestion processes, water purification processes, water supply / distribution processes, petrochemical processes, steel processes, or semiconductor manufacturing processes, multiple process states are measured. Online sensors are installed. Process monitoring equipment (SCADA: Supervisory Control And Data Acquisition) is a process of time-series data (trend graph) that is obtained from process data (flow rate, temperature, water quality, manipulated variable, etc.) obtained by measurement of sensors installed in a normal process system. Convert. This is monitored by a plant manager (manager) or an operator (operator) to grasp the state of the process and change or control the operation of the process. In the time series data of each process data, an upper and lower limit value called a control limit is usually set, and an alarm is issued when the control limit is exceeded. Based on this alarm, the plant manager and operator confirm and review the plant operation. Operation management based on such alarms is fundamental to plant operation.

  Furthermore, in the operation management of the plant that has advanced one step, not only a simple process non-stationary response but also an operation that leads to energy saving and cost saving after achieving a predetermined target performance of the process is required. Here, the predetermined target corresponds to, for example, compliance with the effluent water quality regulations in the case of sewage treatment, and in the case of water purification treatment, the residual chlorine concentration in the purified water is less than the predetermined upper limit or is represented by Cryptosporidium. The target is that no pathogenic microorganisms exist. Further, in the chemical process and the steel process, maintaining the quality (for example, purity and strength) of the product (petroleum refined product and steel) within a predetermined range corresponds to a predetermined target. At this time, it is necessary to monitor the state of the process related to the target performance so as not to fall into the state where the predetermined target is not achieved, and to quickly detect a state change or abnormal state that impedes the achievement of the predetermined target and take measures in advance. This is an important management point. Furthermore, in order to perform operations that lead to energy savings and cost savings while limiting the achievement of predetermined goals, processes that are likely to deviate from the good conditions are always kept in good condition for target performance and energy savings and cost savings. It is necessary to quickly detect a change in state.

  Multivariate statistical process monitoring (MSPC: Multi-Variate) using the “multivariate statistical analysis method” that has been used in the fields of petrochemical processes and steel plants as a method of diagnosing such process changes and abnormalities A method called Statistical Process Control is known.

  As methods that are most often used in MSPC, principal component analysis (PCA) and latent variable projection / partial least squares (PLS) are known.

  In MSPC, using multivariate analysis such as PCA and PLS, (1) detection of plant abnormal signs (Fault Detection) and (2) estimation of process variables (data) that cause abnormalities (Fault Isolation) The main purpose is to do.

  Regarding (1) above, by using the correlation information of a plurality of process variables, it is possible to detect a sign of a minor abnormality that cannot be detected by one variable.

The above (2), after detecting an abnormality in the synthesis anomaly detection data from a plurality of process variables (Q statistic and Hotelling (called T ² statistic Hotelling)), each process for the abnormality detection data A process variable (data) that is a candidate for an abnormal factor is estimated by using a contribution amount that represents the degree of data contribution.

  In this way, when using MSPC, it is compared with conventional monitoring based on simple control limits for individual process variables (in production line monitoring, it is sometimes called SPC (Statistical Process Control) compared to MPSC). Thus, advanced monitoring / diagnosis more useful for plant managers and operators can be performed.

  On the other hand, users such as plant managers and operators should not only take (1) detection of abnormal signs and (2) abnormal factor variable estimation, but also what measures they should take when something abnormal occurs in the process. There is a further strong potential need for countermeasure support information. This need is particularly strong when the plant operator is unskilled. Information such as “there are signs of abnormality” or “diagnostic information indicating that the process variables indicating the abnormality are OO and △ Δ”. Insufficient, seeking support information in the form of “It is recommended to take a measure of □□, because abnormal signs are recognized and the variables that are considered to be XX and △△.” Yes.

JP 2008-59270 A

  There is a large gap between the need for support information and the FDIs (1) and (2) above. The reason is that FDI can make a diagnosis mechanically to some extent (automatically without engineering) without relying on the target process using “statistical methods such as multivariate analysis”. On the other hand, in providing countermeasure support information, it is generally considered that knowledge of the target process and engineering using it are necessary.

  Anomaly cause estimation methods for providing countermeasure support information are: [1] Empirical methods (decision tree, pattern learning, etc.), [2] Logical methods (physical model, fault tree, etc.), [3] Knowledge engineering It is often categorized as a method (expert system, know-how extraction), but any method requires knowledge of the target process in some form or knowledge of the operator who operates the target process. To do. As a result, in order to investigate the true cause of the unsteady state (abnormal state) of the process and to provide countermeasure support, a great deal of engineering is required to build a diagnostic system for each target process. The necessity of development depending on the characteristics of the target process is considered to be an important impediment to hindering the progress of countermeasure support system development during non-stationary times.

  The problem to be solved by the present invention is to provide a process monitoring / diagnosis / support apparatus capable of providing countermeasure support information at the time of abnormality without depending on a target process.

  The process monitoring / diagnosis support apparatus according to the embodiment includes a data collection / save means, a process variable classification means, a process variable correlation definition means, an abnormality detection data definition means, an abnormality detection data contribution amount definition means, an abnormality countermeasure support information definition means , An abnormality detection means, an abnormality factor variable separation means, and an abnormality countermeasure support information provision means.

  The data collection and storage means collects and stores time-series data of n process variables indicating measurement results from n (however, n ≧ 2) sensors that measure the state quantity or operation quantity of the target process.

  The process variable classifying means divides the n process variables into p output variables Y representing the performance index and the performance index, and L representing the manipulated variable (where 1 ≦ p <n). However, classification information indicating the result of classification into the input variable U of 1 ≦ L <n) and the m (where 0 ≦ m <n) intermediate variables Z representing the state quantity and representing the management / monitoring value. Remember.

  The process variable correlation defining means includes correlation information indicating correlation between the n process variables including the classification result from the classification information and time series data of the n process variables over a predetermined period. Define.

  The anomaly detection data defining means generates q anomaly (1 ≦ q << n) anomaly detection data less than the n from time-series data over a predetermined period of the n process variables. An expression and a criterion for determining the presence / absence of an abnormality or an abnormality intensity based on the abnormality detection data are defined.

  The abnormality detection data contribution amount defining means defines a contribution amount of each of the time series data of the n process variables to the abnormality detection data.

  The abnormality countermeasure support information defining means includes a first rule for extracting a predetermined higher-order abnormality factor candidate variable among abnormality factor candidate variables that are process variables that are abnormality factor candidates at the time of abnormality, and the extracted abnormality factor A countermeasure including a second rule for classifying candidate variables into the output variable Y, the input variable U, and the intermediate variable Z based on the correlation information, and a message for assisting countermeasures in case of abnormality according to the classified result A third rule that provides support information is defined.

  The anomaly detection means stores the time-series data of the n process variables at the stored predetermined time and the abnormality detection data corresponding to the time-series data at the predetermined time according to the definition of the abnormality detection data definition means. Based on this abnormality detection data, the presence or absence of abnormality of the data at the predetermined time is diagnosed.

  The abnormality factor variable separation means calculates the contribution amount of the time series data at the predetermined time based on the definition of the contribution amount when there is an abnormality as a result of the diagnosis, and calculates the calculated contribution amount and the first contribution amount. Based on the rule, the abnormal factor candidate variable is separated and extracted from the time-series data at the predetermined time.

  The abnormality countermeasure support information providing means provides the countermeasure support information based on the separated and extracted abnormality factor candidate variable, the correlation information, the second rule, and the third rule.

1 is a schematic diagram showing a basic system configuration of a process monitoring / diagnosis / support apparatus according to a first embodiment. It is a schematic diagram which shows the example of a process variable isolation | separation user interface. It is a schematic diagram which shows Example 1 of a correlation determination rule. It is a schematic diagram which shows Example 2 of a correlation determination rule. It is a schematic diagram which shows Example 3 of a correlation determination rule. It is a schematic diagram which shows the example of a process variable correlation definition interface part. It is a schematic diagram which shows an abnormal factor candidate classification rule. It is a schematic diagram which shows an example of an abnormality countermeasure support message generation rule. It is a schematic diagram which shows an example of an abnormality countermeasure support message generation rule. It is a schematic diagram for demonstrating the effect | action which an abnormality countermeasure assistance message is produced | generated when abnormality occurs. It is a schematic diagram which shows an example of a countermeasure support message. It is a schematic diagram which shows an example of the system configuration | structure in other embodiment. It is a schematic diagram which shows an example of the system configuration | structure in other embodiment. It is a schematic diagram which shows an example of the system configuration | structure in other embodiment.

  Hereinafter, each embodiment will be described with reference to the drawings. Note that the following process monitoring / diagnosis / support apparatus can be implemented with either a hardware configuration or a combined configuration of hardware resources and software. As the software of the combined configuration, a program that is installed in advance from a network or a storage medium into a computer serving as a process monitoring / diagnosis / support apparatus and that realizes the function of the process monitoring / diagnosis / support apparatus is used.

<First Embodiment>
FIG. 1 is a schematic diagram showing a basic configuration of a process monitoring system to which the process monitoring / diagnosis / support apparatus according to the first embodiment is applied.

  This process monitoring system uses, for example, n (however, n ≧ 2) sensors that measure the state quantity or manipulated variable of the target process, with the advanced sewage treatment process 1 for the purpose of removing nitrogen and phosphorus as the target process. It has a process monitoring / diagnosis / support device including a process measurement data collection / storage unit 2 that collects and stores time series data of n process variables indicating measurement results. Although this embodiment does not depend on the target process for that purpose, in order to clarify the implementation image, the present embodiment will be described for the sewage treatment process. In other words, implementation for the sewage treatment process is not an essential constraint.

  The advanced sewage treatment process 1 includes an initial sedimentation basin 101, an anaerobic tank 102, an anaerobic tank 103, an aerobic tank 104, and a final sedimentation basin 105. Further, the advanced sewage treatment process 1 includes an initial settling pond excess sludge removal pump and its extraction flow sensor 111, a blower for supplying oxygen to the aerobic tank 104 and its supply air flow sensor 112, a circulation pump and its circulation flow sensor 113. Each of the return sludge pump and its return flow rate sensor 114, and the final sedimentation basin excess sludge pulling pump and its extraction flow rate sensor 115 has an actuator and its operation amount sensor.

  Further, the sewage altitude treatment system 1 includes a rain sensor 121, a sewage inflow sensor 122 for measuring the inflow sewage amount, an inflow TN sensor 123 for measuring the total nitrogen amount contained in the inflow sewage, and a total phosphorus amount contained in the inflow sewage. An inflow TP sensor 124 to measure, an inflow UV sensor or inflow COD sensor 125 to measure the amount of organic matter contained in the inflow sewage, an anaerobic tank ORP sensor 126 to measure the ORP of the anaerobic tank 102, and an anaerobic tank to measure the pH of the anaerobic tank 102 PH sensor 127, oxygen-free tank ORP sensor 128 that measures the ORP of the oxygen-free tank 103, oxygen-free tank pH sensor 129 that measures the pH of the oxygen-free tank 103, and phosphate sensor that measures the phosphate concentration in the aerobic tank 104 1210, DO sensor 1211 that measures the dissolved oxygen concentration in the aerobic tank 104, and ammonia concentration in the aerobic tank 104 An ammonia sensor 1212, an MLSS sensor 1213 that measures the amount of activated sludge in at least one of the reaction tanks 102 to 104, a water temperature sensor 1214 that measures the water temperature in at least one of the reaction tanks 102 to 104, and the final sedimentation tank 105 Excess sludge SS sensor 1215 that measures the solid matter concentration of the sludge that is pulled from the bottom, the discharge SS sensor 1216 that measures the SS concentration of the effluent discharged from the final sedimentation basin 105, and the sludge interface level of the final sedimentation basin 105 Sludge interface sensor 1217, sewage discharge sensor 1218 for measuring the amount of discharged sewage, discharge TN sensor 1219 for measuring the total amount of nitrogen contained in the discharged sewage, discharge TP sensor 1220 for measuring the total amount of phosphorus contained in the discharged sewage, And a discharge UV sensor or discharge CO for measuring the amount of organic matter contained in the discharged sewage Each sensor 1221, having a process sensor.

  Here, the various actuators 111 to 115 described above operate at a predetermined cycle. Further, the operation amount sensor group of the various actuators 111 to 115 and the various process sensors 121 to 1221 perform measurement at a predetermined cycle.

  The process monitoring / diagnosis / support apparatus collects time series data (process data) of process variables obtained at predetermined intervals from the operation amount sensor groups of the various actuators 111 to 115 and the various process sensors 121 to 1221 in the memory. It has a process measurement data collection / storage unit 2 for holding it.

  Further, the process monitoring / diagnosis / support apparatus includes a process variable registration unit 3, a process variable classification user interface unit 4, an abnormality diagnosis model construction unit 5, a process variable correlation definition interface unit 6, an abnormality monitoring / diagnosis / measure support unit 7, An abnormality countermeasure record holding unit 8 and a user interface unit 9 are provided.

  The process variable registration unit 3 constructs the process monitoring / diagnosis / support apparatus of the present embodiment from among the operation variable sensors of the various actuators 111 to 115 and the process variables measured by the various process sensors 121 to 1221. Register the name of the process variable required to do this.

  The process variable classification user interface unit 4 displays a list of process variables registered by the process variable registration unit 3 and classifies the process variables therein into an input variable U, an output variable Y, and an intermediate variable X. It is. Supplementally, the process variable classification user interface unit 4 displays the names of n process variables, and during this display, the input variables, output variables, and intermediate variables in the process variable classification unit 51 are displayed according to the user's operation. select.

  The abnormality diagnosis model construction unit 5 includes a past time series of process variables registered by the process variable registration unit 3 among the process variable measured by the operation amount sensors of the various actuators 111 to 115 and the various process sensors 121 to 1221. Data is extracted from the process measurement data collection / storage unit 2 to construct an abnormality diagnosis model.

  The process variable correlation definition interface unit 6 is a user interface that presents correlation information between process variables defined by the abnormality diagnosis model construction unit 5 to the user, and allows the user to correct the information.

  The abnormality monitoring / diagnosis / countermeasure support unit 7 presents the process variables registered by the process variable registration unit 3 among the operation variable sensors of the various actuators 111 to 115 and the process variables measured by the various process sensors 121 to 1221. Time series data is extracted from the process measurement data collection and storage unit 2, and the function defined by the abnormality diagnosis model construction unit 5 is used to detect abnormal signs, extract the cause variable candidates, and provide support information for abnormality countermeasures. To do.

  The abnormality countermeasure record holding unit 8 holds a result of countermeasures actually taken based on information from the abnormality monitoring / diagnosis / countermeasure support unit 7. Supplementally, the abnormality countermeasure record holding unit 8 records the operation performed by the user after presenting the countermeasure support information provided by the abnormality countermeasure support information providing unit 74. Here, when the countermeasure support information is provided, the abnormality countermeasure record holding unit 8 determines the values of the input variable U, the output variable Y, and the intermediate variable Z before the current time based on the time series data and the classification information. The value and the countermeasure support information are presented. Further, after presenting the countermeasure support information, the abnormality countermeasure record holding unit 8 continuously records the operation record of the input variable U, the value of the output variable Y, and the value of the intermediate variable Z as an operation performed by the user. Record.

  The user interface unit 9 includes a process variable registration unit 3, a process variable classification user interface unit 4, and a process variable correlation definition interface unit 6, and notifies the plant manager and operator of information from the abnormality monitoring / diagnosis / countermeasure support unit 7. Then, a countermeasure instruction based on the countermeasure support information can be sent to the plant as a command signal, and a recording operation instruction can be sent to the abnormality countermeasure record holding unit 8.

  The abnormality diagnosis model construction unit 5 includes a process variable classification unit 51, a past (offline) data extraction unit 52, a process variable correlation definition unit 53, an abnormality detection data definition unit 54, an abnormality detection data contribution amount definition unit 55, and an abnormality time. A countermeasure support information definition unit 56 is provided.

  As shown in FIG. 2, the process variable classification unit 51 classifies process variables into input variables, output variables, and intermediate variables according to the input of the process variable classification user interface unit 4 and supplies the information at the time of countermeasure against abnormality. Supplementally, the process variable classifying unit 51 divides n process variables into p (where 1 ≦ p <n) output variables Y that are state quantities and performance indicators, and L ( However, classification information indicating the result of classification into input variables U of 1 ≦ L <n) and m intermediate variables Z (0 ≦ m <n) which are state quantities and represent management / monitoring values. Has a function to remember.

  The past (offline) data extraction unit 52 extracts past offline data of variables registered by the process variable registration unit 3 from the process measurement data collection / storage unit 2.

  The process variable correlation definition unit 53 uses the offline data extracted by the past (offline) data extraction unit 52 to define correlation information between the process variables defined by the process variable registration unit 3. Supplementally, the process variable correlation definition unit 53 includes the classification result indicated by the classification information as shown in FIGS. 3 to 6 from the time series data over a predetermined period of the classification information and n process variables. Correlation information indicating the correlation among the n process variables is defined. Note that the process variable correlation definition unit 53 may define correlation information based on a correlation matrix or principal component load matrix of time-series data over a predetermined period of n process variables. Further, the process variable correlation definition unit 53 may include a table of n rows and n columns representing the presence or absence or strength of correlation between n process variables as correlation information. In this case, the process variable correlation definition interface unit 6 displays this correlation information table and corrects the contents of the table in accordance with the user's operation.

  The abnormality detection data definition unit 54 uses the offline data extracted by the past (offline) data extraction unit 52 to define a method for synthesizing the abnormality detection data. Supplementally, the abnormality detection data definition unit 54 generates q pieces of abnormality detection data less than n (where 1 ≦ q << n) from time-series data over a predetermined period of n process variables. And a criterion for determining the presence or absence of abnormality or the strength of abnormality based on the abnormality detection data.

  Here, the anomaly detection data definition unit 54 performs principal component analysis (PCA), latent variable projection (PLS), principal component regression (PCR), canonical correlation analysis (CVA), Mahalanobis distance, and robust principal component analysis ( Robust PCA), kernel principal component analysis (Kernel PCA), discriminant analysis, fuzzy C-means clustering, K-means clustering, support vector machine (SVM), relevance vector machine (RVM), and subspace method You may define the formula which produces | generates the data for abnormality detection using two or more methods.

  The anomaly detection data contribution amount definition unit 55 is a process variable contribution amount registered in the process variable registration unit 3 with respect to the anomaly detection data defined by the anomaly detection data definition unit 54 (time series of n process variables). Each contribution of data). The anomaly detection data contribution amount definition unit 55 calculates the contribution amount of the target process variable k among the n process variables (where k = 1, 2,..., N) from the following [a] to You may define by either of [c].

[A] It is defined by projecting from the anomaly detection data defined by the anomaly detection data defining unit 54 to the space defined by the process variable k.
[B] The abnormality detection data is decomposed into the sum of n components and defined as the k-th component that only the process variable k affects.
[C] Define using independent component analysis.

  The abnormality countermeasure support information definition unit 56 includes classification information of input variables, output variables, and intermediate variables from the process variable classification unit 51, correlation information between process variables defined by the process variable correlation definition unit 53, and abnormality detection. A rule for providing support information for advising a countermeasure method at the time of abnormality is defined from the information regarding the contribution amount defined by the data contribution amount definition unit 55.

  Supplementally, the abnormality countermeasure support information defining unit 56 extracts a first upper-order abnormality factor candidate variable from among abnormality factor candidate variables that are process variables that are abnormality factor candidates at the time of abnormality, and FIG. As shown in FIG. 8 and FIG. 9, the second rule for classifying the extracted abnormal factor candidate variables into the output variable Y, the input variable U, and the intermediate variable Z based on the correlation information is classified. A third rule that provides countermeasure support information including a message that supports countermeasures in case of abnormality is defined according to the result.

  Here, as the second rule, the extracted abnormal factor candidate variables are classified into the input factor variable FU belonging to the input variable U, the output factor variable FY belonging to the output variable Y, and the intermediate belonging to the intermediate variable Z based on the classification information. A rule for classification into the factor variable FZ may be included.

  The third rule may include a fourth rule A, a fifth rule B, and a sixth rule C.

  When the fourth rule A includes the output factor variable FY in the abnormal factor candidate variable, the fourth rule A notifies that the output factor variable FY has deteriorated and is correlated with the output factor variable FY by the correlation information. It provides that countermeasure support information including a message that prompts the user to confirm and review the value of the defined input variable U is provided.

  The fourth rule A stipulates that when the input factor variable FU is included in the input variable U, countermeasure support information including a message that prompts the user to review the input factor variable FU is provided. It is preferable to do.

  Further, as the fourth rule A, the intermediate variable Z (generally a plurality) defined as correlated with the output factor variable FY by the correlation information is managed, and the intermediate factor Z includes the intermediate factor variable FZ. More preferably, the provision of countermeasure support information including a message prompting the management of the value of the intermediate factor variable FZ to return to a normal value is provided.

  The fifth rule B is an abnormal state in which the input factor variable FU deviates from the normal time when the abnormal cause candidate variable does not include the output factor variable FY but includes the input factor variable FU. The countermeasure support information including a message notifying that there is a possibility that the output variable Y defined as having correlation with the input factor variable FU by the correlation information may deteriorate in the future. It stipulates that it will be provided.

  Further, as the fifth rule B, in addition to this rule, the intermediate variable Z defined as correlated with the input factor variable FU by the correlation information may be deteriorated in the future. In the case where the intermediate factor variable FZ is included, it is further defined that provision of countermeasure support information including a message notifying that the deterioration factor may be caused by the input factor variable FU is provided. preferable.

  When the abnormal factor variable is only the intermediate factor variable FZ, the sixth rule C notifies that the intermediate factor variable FZ has deteriorated and is correlated with the intermediate factor variable FZ by the correlation information. It defines that countermeasure support information including a message notifying that the defined output variable Y may deteriorate in the future is provided.

  Further, as the sixth rule C, in addition to this rule, the value of the input variable U that is defined as correlated with the intermediate factor variable FZ is confirmed by the correlation information, and the necessity of changing the operation is examined. It is further preferable to define that the message is included.

  On the other hand, the abnormality monitoring / diagnosis / countermeasure support unit 7 includes a current (online) data extraction unit 71, an abnormality detection unit 72, an abnormality factor variable separation unit 73, and an abnormality countermeasure support information provision unit 74.

  The current (online) data extraction unit 71 extracts the current online data of the variable registered by the process variable registration unit 3 from the various time series data stored in the process measurement data collection / storage unit 2.

  The anomaly detector 72 generates the anomaly detection data by inputting the online data extracted by the current (online) data extractor 71 into the anomaly detection data calculation formula defined by the anomaly detection data definition unit 54 Thus, the current abnormality degree is checked, and the presence or absence of abnormality is detected in accordance with a preset abnormality / normality determination criterion. Supplementally, the abnormality detection unit 72 stores the time-series data of the n process variables stored at a predetermined time and the abnormality detection data corresponding to the time-series data at the predetermined time according to the definition of the abnormality detection data definition unit. Based on this abnormality detection data, the presence / absence of abnormality of the data at the predetermined time is detected (diagnosed).

  When an abnormality is detected by the abnormality detection unit 72, the abnormality factor variable separation unit 73 calculates the degree of contribution of each process variable defined by the abnormality detection data contribution amount definition unit 55 to the abnormality. Supplementally, the abnormality factor variable separation unit 73 calculates the contribution amount of the time series data at the predetermined time based on the definition of the contribution amount when there is an abnormality as a result of the detection (diagnosis) by the abnormality detection unit 72, Based on the calculated contribution amount and the first rule of the abnormality countermeasure support information definition unit 56, the abnormal factor candidate variable is separated and extracted from the time-series data at the predetermined time.

  As shown in FIGS. 10 and 11, the abnormality countermeasure support information providing unit 74 includes information on process variables, which are predetermined higher-order abnormality factor candidates extracted by the abnormality factor variable separation unit 73, correlation information, According to the second rule and the third rule defined by the countermeasure support information definition unit 56, the countermeasure support information at the time of abnormality is provided to the operator through the user interface unit 9.

  Next, the operation of the process monitoring system configured as described above will be described.

  First, in the sewage altitude treatment process 1, process information is measured at a predetermined cycle by the operation amount sensor group of the various actuators 111 to 115 and the various process sensors 121 to 1221. These pieces of measurement information are stored as time series data by the process measurement data collection / storage unit 2 according to a predetermined format.

  When constructing a process monitoring / diagnosis / support apparatus, first, the process variable registration unit 3 stores the process measurement data collection / storage unit 2, the operation amount sensor groups of the various actuators 111 to 115, and the various process sensors. Of the items measured in 121 to 1211, which variable is used when configuring the process monitoring / diagnosis / support apparatus is defined.

  For example, “[1] Air supply amount” corresponding to the blower for supplying oxygen to the aerobic tank 104 and its supply air flow rate sensor 112, “[2] Circulation flow rate” corresponding to the circulation pump and its circulation flow rate sensor 113, "[3] Return flow rate" corresponding to the return sludge pump and its return flow sensor 114, "[4] Surplus flow rate" corresponding to the final sedimentation basin excess sludge pulling pump and its extraction flow sensor 115, corresponding to the rain sensor 121 “[5] Rainfall”, “[6] Sewage inflow” corresponding to the sewage inflow sensor 122 that measures the inflow sewage, “corresponding to the inflow TN sensor 123 that measures the total amount of nitrogen contained in the inflow sewage” [7] Inflow TN ”,“ [8] Inflow TP ”corresponding to the inflow TP sensor 124 that measures the total amount of phosphorus contained in the inflow sewage, and the amount of organic matter contained in the inflow sewage “[9] Inflow UV” corresponding to the inflow UV sensor or the inflow COD sensor 125, “[10] Anaerobic tank ORP” corresponding to the anaerobic tank ORP sensor 126 for measuring the ORP of the anaerobic tank 102, pH of the anaerobic tank 102 "[11] Anaerobic tank pH" corresponding to the anaerobic tank PH sensor 127 for measuring the oxygen, "[12] Anoxic tank ORP", corresponding to the anaerobic tank ORP sensor 128 for measuring the ORP of the oxygen-free tank 103, oxygen-free “[13] Oxygen tank pH” corresponding to the anaerobic tank pH sensor 129 for measuring the pH of the tank 103, “[14] Aerobic corresponding to the phosphate sensor 1210 for measuring the phosphoric acid concentration in the aerobic tank 104. PO4-P "," [15] Aerobic tank DO "corresponding to the DO sensor 1211 that measures the dissolved oxygen concentration in the aerobic tank 104, Ammoni that measures the ammonia concentration in the aerobic tank 104 "[16] Aerobic tank ML4-N" corresponding to the sensor 1212 and "[17] Aerobic tank MLSS" corresponding to the MLSS sensor 1213 that measures the amount of activated sludge in at least one of the reaction tanks 102 to 104. "[18] Reaction tank water temperature" corresponding to the water temperature sensor 1214 for measuring the water temperature in at least one of the reaction tanks 102 to 104, corresponding to the discharge TN sensor 1219 for measuring the total amount of nitrogen contained in the discharged sewage. “[19] Discharge TN” is registered as a variable using “[20] Discharge TP” corresponding to the discharge TP sensor 1220 for measuring the total amount of phosphorus contained in the discharged sewage. It is preferable to construct a registration screen as a GUI on the monitoring screen so that the user can easily perform this registration.

  The function of registering such process variables is the function of the process variable registration unit 3.

  Next, the process variable classification user interface unit 4 displays a list of registered process variables as shown in FIG. 2 as a GUI on the monitoring screen, and among them, an input variable, an output variable, and an intermediate variable can be selected. Keep it like that. In FIG. 2, the input variable, output variable, and intermediate variable fields are indicated by a circle. However, when an actual monitoring screen (GUI) is constructed, a check box is used for checking. Also, it is more efficient to save the labor of checking if the default state is set as an intermediate variable and only the input variable and the output variable can be selected. Such a GUI is an example of a process classification user interface corresponding to claim 2 and an example of the operation of the process variable classification user interface unit 4.

  Next, in the process variable classification unit 51, the registered process variable is converted into an input variable that represents the manipulated variable of the process according to the criteria that the variable registered in the process variable registration unit 3 is checked in the process variable classification user interface unit 4. There are three categories: output variables representing process performance indicators and intermediate variables corresponding to process management values.

  The above example is classified as follows, for example.

  Input variables: [1] air supply amount, [2] circulation flow rate, [3] return flow rate, [4] surplus flow rate.

  Output variables: [19] discharge TN, [20] discharge TP.

  Intermediate variables: [5] Rainfall, [6] Sewage inflow, [7] Inflow TN, [8] Inflow TP, [9] Inflow UV, [10], Anaerobic tank ORP, [11] Anaerobic tank pH, [12 ] Oxygen tank ORP, [13] Oxygen tank pH, [14] Aerobic tank PO4-P, [15] Aerobic tank DO, [16] Aerobic tank NH4-N, [17] Aerobic tank MLSS, [18] Reaction vessel water temperature.

  Input variables are variables that can be changed directly by operating a blower (blower) or pump. Besides these, the amount of pumped water (pumped water) that pumps water into the injection amount of chemicals and the sewage treatment process Pump) etc. can also be input variables.

  Since the direct purpose of the sewage treatment is to maintain the discharged water quality within the regulation value range, the output variable is the output variable as described above. In addition to the above, if there is effluent quality to be monitored, it can also be an output variable. Furthermore, the discharge water quality load amount obtained by multiplying the discharge water quality by the discharge water amount can be an output variable. Furthermore, when the amount of power is monitored, the amount of power consumption or the energy intensity obtained by dividing the amount of power by the processing amount can also be an output variable.

  As shown above, the intermediate variable represents the state of the process and the disturbance element input to the process, and is converted using these as well as the direct management / monitoring values as described above. Management values may be defined as intermediate variables. Typical examples include control values specific to sewage treatment processes such as HRT (hydraulic residence time), SRT (sludge residence time), A-SRT (aerobic tank sludge residence time), or BOS-SS load. is there. These may be included as intermediate variables by incorporating formulas that pre-calculate these values.

  Although there is no objective standard for classification of intermediate variables and output variables, it is somewhat subjective, but it may be defined as appropriate at the judgment of the parties who construct this process monitoring / diagnosis / support apparatus.

  This is the function of the process variable classification unit 51.

  Next, in the past (offline) data extraction unit 52, the past over a predetermined period of the variable registered in the process variable registration unit 3 from the past process data stored in the process measurement data collection / storage unit 2. Extract the data. At this time, it is preferable that the “predetermined period” is implemented as a GUI on the monitoring screen together with the process variable registration unit 3 so that the user can easily select as appropriate. Hereinafter, the past time series data extracted here will be described as X. This X is a matrix having variables in the row direction (variables [1] to [20] in the above case) and time series samples (time series data) over the “predetermined period” defined above in the column direction. In the description, the number of variables is n and the number of time series data is m. Therefore, X is m × n time-series data. This is the operation of the past (offline) data extraction unit 52.

  Next, the process variable correlation definition unit 53 removes missing data and outliers from the X extracted by the past (offline) data extraction unit 52, and a plurality of monitoring items (monitoring model configuration variables) having different physical dimensions. The correlation matrix of X is obtained after appropriately performing normalization (= subtracting the average value and dividing by the variance). At this time, if necessary, decimation (thinning time-series samples at a predetermined period) may be performed. In this case, the number m of time-series data changes. The data processed in (1) is described again as X. The correlation matrix of X matches the variance-covariance matrix when each variable is normalized, and can be simply calculated by the following formula.

Correlation matrix S = X ^T X (1)
Expression (1) represents a matrix of the number of process variables n × n, and each element thereof is a real number representing a correlation coefficient from −1 to 1. Further, from the definition of the correlation coefficient, the diagonal elements are all 1 (the correlation coefficient with itself is 1) and the symmetric matrix (the correlation between a and b is the same as the correlation between b and a). Appropriate criteria are set for the correlation coefficient of the equation (1), and the presence or strength of correlation is defined. The simplest definition is based on the presence or absence of correlation. For example, a correlation determination rule (example 1) can be defined as shown in FIG.

  Here, Th is a reference for determining the presence or absence of correlation, and for example, Th = 0.5 or Th = 0.7 (Th2≈0.5) may be used. Also, an appropriate expression format is used, such as “correlation = 1”, “no correlation = 0”, “correlation = O”, and “no correlation = x”. SS is a matrix indicating the presence or absence of correlation, and this is an example of a matrix defined by the process variable correlation definition unit 53.

  As another example, the strength of the correlation is determined in several steps. For example, in the case of dividing into three steps, a correlation determination rule (example 2) is set as shown in FIG. Also good.

  Here, Th1 and Th2 are thresholds satisfying 0 <Th2 <Th1 <1, and may be set as Th1 = 0.7, Th2 = 0.5, for example. Also, an appropriate expression format can be used such as strong correlation = 2, weak correlation = 1, no correlation = 0, strong correlation =＝, weak correlation = ◯, no correlation = ×.

Further, not only from the viewpoint of correlation between variables, but also the presence or absence (or strength) of the relationship between the variables can be defined using a loading matrix for principal component analysis described later. The size n × n principal component loading considering all variables and P _a, will be described the k-th row and j-th column P _a (k, j) and.

P _a (k, j) has a property that the loading matrix P _a becomes an orthonormal matrix, and its elements take values in the range from −1 to 1, and are defined by k rows of P _a . The square sum Σ_ {j = 1} ^ {15} P _a (k, j) ² of the i principal component (hereinafter referred to as P _a (k, :)) is 1. This means that the average value of P _a (k, j) ² is 1 / n. If this property is used, the presence or absence of the relationship between each variable can be determined by a correlation determination rule (example 3) as shown in FIG.

  In the above, the relationship is defined based on the criterion that the components larger than the average value of the k-th principal component shown in the k-th row of the principal component loading matrix (principal component loading matrix) are related to each other. is there. Of course, the average value 1 / n can be appropriately modified to have different threshold values, or the strength of the relationship can be defined by a plurality of threshold values as in the correlation determination rule example 2.

  FIG. 6 shows an example in which the strength of the relationship (or correlation) between process variables is defined in this way. In the process variable correlation definition section 53, the correlation between the process variables is finally defined as shown in FIG. 6 by some mechanical means. At this time, it should be noted that the input variables, intermediate variables, and output variables classified by the process variable classification unit 51 are classified and held as shown in FIG. It becomes a point when providing countermeasure support.

  The above embodiment is an example of the process variable defining means corresponding to claim 5. The above is the operation of the process variable correlation definition unit 53.

  At this time, it is desirable that a table showing the correlation between the process variables as shown in FIG. 6 is presented to the user through the GUI on the monitoring screen. In this case, the process variable correlation definition interface unit 6 allows the user to appropriately rewrite the presence / absence and strength of the correlation in this table. Then, when the user feels a doubt and wants to correct the correlation determined mechanically as described above, the user's request can be easily reflected in the system. In this case, it does not require the user to define the presence or absence of correlation in advance, but it is configured to enable it when the user wants to customize the correlation defined mechanically Is the point of this embodiment.

  This action is an example of a user interface corresponding to claim 6 and an action example of the process variable correlation definition interface unit 6.

  Next, the abnormality detection data definition unit 54 generates a small number of abnormality detection data by applying various methods of multivariate analysis and machine learning to X extracted by the past (offline) data extraction unit 52. . What is important here is that the abnormality detection data is generated from X, and therefore includes information on n process variables, and the abnormality detection data is much smaller than n and about 1-2. It is a small number.

As such a method, a method often used as a process diagnosis technique is called MSPC (multivariate statistical process management), and usually uses principal component analysis (PCA) or latent variable projection (PLS). Anomaly detection data called Q statistics and T ² statistics are synthesized.

  PCA treats all n variables as equivalent variables, whereas PLS treats n variables separately as input variables and output variables.

  In this embodiment, process variables are classified into input variables, output variables, and intermediate variables. However, this is classified for use in support of abnormality countermeasures after diagnosis, regardless of this classification. Either PCA or PLS can be used. When applying PCA, input variables, output variables, and intermediate variables that have already been defined may be applied without any distinction. When PLS is used, it is applied by reassigning the classified intermediate variables to input variables or output variables. Instead of PLS, PCR combining PCA and multiple regression analysis can be used. In addition, when a plurality of process variables are repeatedly measured at different places, canonical correlation analysis (CVA) can be used. As such an example, there may be mentioned a case where five variables such as water temperature, pH, conductivity, nitric acid, and turbidity are repeatedly measured at a plurality of locations in the water treatment process.

  In addition, when it is assumed that a large amount of outliers are included in the data, for example, Non-Patent Document 1 “Mia Hubert, Peter J. Rousseeuw, Karlien V,“ ROBPCA: a New Approach to Robust Principal Component Analysis (2005) “Technometrics” and non-patent document 2 “C Croux, A Ruiz-Gazen, High breakdown estimators for principal components: the projection-pursuit approach revisited, Journal of Multivariate Analysis” Various robust PCA algorithms may be used. Alternatively, this may be expanded and used as a robust PLS.

  Furthermore, when a strong nonlinear correlation is assumed between the data, for example, Non-Patent Document 3 “K.-R. Muller, S. Mika, G. Ratsch, K. Tsuda, and B. Scholkopf, An Introduction to kernel-based learning algorithms. IEEE Trans. Pattern Anal. Machine Intell., 12 (2): 181.201, March 2001. ”and Non-Patent Document 4“ B. Scholkopf, AJ Smola, and K.-R. Muller , Nonlinear component analysis as a kernel eigenvalue problem. Neural Computation, 10 (5): 1299.1319, 1998. Also good. Furthermore, when both nonlinearity and outlier problems exist, a method combining a robust PCA and a kernel PCA can be used.

Further, as a technique similar to MSPC, detection data can be generated using Mahalanobis distance such as Taguchi method used in the field of quality engineering. The detection data using the Mahalanobis distance is of T ² statistic Hotelling essentially equivalent using PCA to be described later, in the method using the PCA, is casting the dimension to low dimension by PCA The point is different. However, even the method using the Mahalanobis distance may be reduced in order for numerical stabilization, and is essentially the same as the T ² statistic.

  In addition, by treating abnormality detection as a two-class clustering problem that identifies certain data as a normal class and an abnormal class, various clustering techniques used in the field of machine learning can be used. As such, clustering such as discriminant analysis, K-means clustering, fuzzy C-means clustering may be used, and support vector machines (SVM) and relevance vector machines (RVM) using Bayesian inference are used. You may apply as 2 classes of clustering. When using these clustering, it is necessary to define the data for anomaly detection appropriately, but the simplest method is to define the distance from the center of the normal class separated by clustering as the data for anomaly detection. Can do.

  As another technique in the field of machine learning, a technique of subspace method often used in the field of image recognition can be applied. At this time, the concept of “similarity” defined by the subspace method can be used as abnormality detection data. In other words, data having a degree of similarity far from normal data can be defined and used as abnormal data. However, the concept of similarity is a concept very close to the Q statistic by PCA, which will be described later, and the Q statistic represents an “abnormality” indicating a distance from data to be diagnosed to a low-dimensional space. On the other hand, the similarity is an amount obtained by projecting data to be diagnosed onto a low-dimensional space, and is a “normality” indicating how close the certain data is to the low-dimensional space. These are amounts of complementary spaces orthogonal to each other and can be converted to each other, and thus are substantially the same in principle.

  In addition, the abnormality detection data can be synthesized by combining the above methods in various forms. For example, a combination of replacing the multiple regression analysis part of PCR combining PCA and multiple regression analysis with support vector regression (SVR), which is an extension of support vector machine to regression analysis, or replacing with RVM applied to regression problems May be performed.

  In any case, it is important for the abnormality detection data definition unit 54 to generate a small number of abnormality detection data from X, and any method may be used.

In the following, a specific calculation formula for generating abnormality detection data when ordinary PCA is used will be described in order to explain the subsequent operation. When PCA is used, X can be decomposed as follows.

T _a ∈R ^{m × n} is a matrix called a score matrix consisting of m samples (or time series data) and n principal components, and P _a ∈R ^{n × n} is an n configuration. It is a matrix called a loading matrix indicating the relationship between variables and n principal components. TεR ^{m × p} is a submatrix of T _a truncated ^{by p} << ⁿ principal components, and is usually called a score matrix. Similarly, P∈R n ^{× p} is a P _a submatrix representing the relationship between the main component has been aborted by p«n number for n variables, typically called the P and loading matrices Yes. E∈R ^{m × n} is an error matrix composed of m samples (or time-series data) and n variables, and represents an error when the principal component is truncated at p << n.

In the following, T _a and T and P _a and P are clearly distinguished, T _a and P _a are called a score matrix and a loading matrix, respectively, and T and P are called a main score matrix and a main loading matrix, respectively. . Using these, the following Q statistics and Hotelling's T ² statistics are defined as anomaly detection data.

Q statistics:
Q (x (t)) = x T (t) (I-PP T) x (t) (3)
Hotelling's T ² statistic:
T ² (x (t)) = x ^T (t) P ^T Λ ⁻¹ Px (t) (4)
Here, Λ is a matrix having the principal component variance as a diagonal element, and means that the variance is normalized. I is a unit matrix of an appropriate size. X (t) is the t-th element of the matrix X. At the time of abnormality monitoring / diagnosis described later, this x (t) is calculated by replacing it with process data measured online. Expressions (3) and (4) are examples of abnormality detection data defined by the abnormality detection data definition unit 54. Here, two abnormality detection data are used.

  Furthermore, a threshold is set as a criterion for discriminating between abnormality and normality for these statistics. Since this threshold setting value is greatly related to the detection of state changes and abnormal signs, its setting method is important, but since it is not related to the idea of this embodiment, only a typical setting method is described. deep.

If, when no prior information is not for past offline data, as the default setting methods, the use of statistical confidence limits about the T ² statistic statistical confidence limits and Hotelling Q statistic (Non-Patent Document 5: C. Rosen “Monitoring Wastewater Treatment Systems”, Lic. Thesis, Dept. of Industrial Electrical Engineering and Automation, Lund University, Lund, Sweden (1998)).

These can be written as follows:

Here, p is the number of variables left in the model. c _α is the deviation of the standard deviation of the standard normal distribution when the limit of the confidence interval is 1-α (eg, when α = 0.01, 2.53, when α = 0.05, 1.96 ). Also, λ _i is a diagonal element of Λ (that is, Θ _i is the sum of i powers of each component included in the error term).

  Where p is the number of variables selected (= remaining in the model) and m is the number of all variables. F (p, mp, α) is an F distribution when the degree of freedom is (p, mp) and the reliability limit is α (= 0.01 or 0.05 in many cases). It is.

  The above is an example of the operation of the abnormality detection data definition unit 54 and corresponds to claim 3.

  Next, the abnormality detection data contribution amount definition unit 55 sets a definition formula for the contribution amount of the input variable of each diagnosis model to the statistic (abnormality detection data) defined by the equations (3) and (4). To do.

  There are a plurality of methods for defining the contribution amount. For example, the contribution amount can be defined as follows.

Q statistics contribution:
Q _cont (n, t) = x ^T (t, n) F (:, n) ^T F (:, n) x (t, n) (7)
F = (I−PP ^T )
The contribution of Hotelling's T ² statistic:
T ² cont (n, t) = x ^T (t) P ^T Λ ⁻¹ P (:, n) x (t, n) (8)
Here, n means the nth variable, and t is a variable representing a certain time. Using the expressions (7) and (8), it is possible to calculate how much each process variable contributes to the value of the abnormality detection data. Equation (7) is a projection of the Q statistic onto the axis of the nth process variable. Equation (8) is not a simple projection, but the T ² statistic is the sum of n variable components. Skillfully disassembled.

When using data that is not the Q statistic or the T ² statistic as the abnormality detection data, it is necessary to appropriately define the contribution amount based on the concept similar to the equations (7) and (8).

  When the Mahalanobis distance used in quality engineering is used, the contribution amount can be defined using a sensitivity analysis method used in the Taguchi method.

When applying the clustering method, if the abnormality detection data is defined as the distance from the center of the normal class as described above, the component to which each process variable contributes to this distance can be defined as the contribution amount. . For example, when defining by the Euclidean distance, if the distance from the center of the normal class to the diagnostic data is D, D ² = D ₁ ² + D ₂ ² + D ₃ ² + ... + D _n ² , D _i ² , i = 1, 2,..., N can be decomposed into the distance of each component of the diagnostic data, and D _i ² can be defined as the contribution amount.

  In addition, after detecting anomalies, use independent component analysis (ICA) to analyze which data is likely to cause the anomaly, and there are some results even if the contribution is not explicitly defined. It may be possible to extract process variables that are candidates for abnormal factors.

  Whatever method is used, the abnormality detection data contribution amount definition unit 55 calculates the abnormality detection data of the diagnostic data calculated according to the method defined by the abnormality detection data definition unit 54 for certain diagnostic data. If you input, you must have a mechanism that can rank and output which process variables are likely to cause abnormalities, and any method can be used if this can be achieved. Absent.

  The above operation corresponds to claim 4, and is an operation example of the abnormality detection data contribution amount definition unit 55.

  Next, the abnormality countermeasure support information definition unit 56 includes process variable information ranked in descending order of the possibility of an abnormality factor supplied as an output of the abnormality detection data contribution amount definition unit 55, and a process variable correlation definition unit. Based on the information from 53, a rule for providing information for supporting countermeasures against abnormality is established.

  First, a high-order factor candidate is extracted from among the ranked abnormal factor candidate variables according to a predetermined rule. The simplest rule is a rule for extracting several top (for example, three) abnormal factor candidates as high-order factor candidates. In this method, only the rank information of the abnormal factor candidates is used, and it is not considered how strong the abnormality is. Therefore, a method of extracting one having a high abnormality is more preferable. For example, the average value μ and the standard deviation σ of the contribution amount may be calculated, and a rule may be set such that a variable whose distance from the average μ exceeds kσ (for example, k = 3) is an abnormal factor candidate. . At this time, μ and σ may be calculated for the diagnosis target data at the time of abnormality detection, or may be calculated in advance from the contribution amount data for a predetermined period. In the former case, abnormal data is always included (because when an abnormal condition is detected), and therefore it is preferable to perform some kind of robust processing such as trim averaging and trim distribution that are performed by discarding some data. Further, the above two rules may be combined and a rule may be adopted in which a maximum of three items having a high degree of abnormality are listed as candidates. In any case, the first rule defined by the abnormality countermeasure support information definition unit 56 is to first extract some higher-order abnormality factor candidates.

  Next, a rule for classifying the extracted abnormal factor candidate variables into input variables, output variables, and intermediate variables shown in FIG. 6 is set. For example, when the definition of extracting the top three variables in the first rule is adopted, classification is made by associating the input variables, output variables, and intermediate variables with the three variables to the table in FIG. To do. This can be easily realized. If the abnormality factor candidate variable (vector) is FX, the input variable (vector) in this is FU, the intermediate variable (vector) is FZ, and the output variable (vector) is FY, it is simply shown in FIG. As shown in FIG. 7, an abnormal factor candidate classification rule may be applied.

  Here, [] represents an empty vector, and [AB] represents a vector obtained by connecting the vector A and the vector B.

  As described above, in the second rule, the abnormal factor candidate variable FX is classified into the input variable FU, the output variable FY, and the intermediate variable FZ.

  Next, a template for providing information for supporting countermeasures is created using these input variable FU, output variable FY, and intermediate variable FZ. This is the third rule. The point here is that the input variable FU, the output variable FY, and the intermediate variable FZ are associated with the correlation matrix in FIG. 6 and a support message is issued. Incorporating a support message about whether it ’s okay. For operators, information on what kind of operation should be performed at the time of abnormality, and what kind of state and how the performance index that the operator attaches importance to may change Since the information is considered to be important information, for example, an abnormality countermeasure support message generation rule template as shown in FIGS. 8 and 9 is created.

  Such a message template does not require any special knowledge about the target plant, simply classifies process variables into input variables, output variables, and intermediate variables, and creates a table that indicates whether there is a relationship between each process variable. The point is that it can be created just by using it. Further, as described above, a table that indicates whether or not there is a relationship between process variables can be automatically generated from a correlation matrix, a principal component loading matrix, or the like. Therefore, if such a message template is created, countermeasure support information at the time of abnormality can be automatically provided according to this template.

  The above embodiment corresponds to claim 7, and is an example of the operation of the abnormality countermeasure support information definition unit 56.

  According to the above procedure, the construction of the abnormality monitoring / diagnosis / countermeasure support model in the abnormality diagnosis model construction unit 5 is completed. Next, the abnormality monitoring / diagnosis / countermeasure support unit 7 is supplied with the abnormality monitoring / diagnosis / countermeasure support model constructed by the abnormality diagnosis model construction unit 5 and uses this model to detect an abnormality, estimate a factor variable, and Provide countermeasure support information.

  First, the current data (online data) extraction unit 71 extracts online data at a time point (hereinafter referred to as the present time or the present time) at which the diagnosis is collected by the process measurement data collection / storage unit 2. Using the current data extracted by the current (online) data extraction unit 71, the abnormality monitoring / diagnosis / countermeasure support unit 7 monitors the process state, and when there is a change in the state or a sign of abnormality is recognized. Detects it.

Next, the anomaly detection unit 72 first extracts the current data corresponding to the variable registered by the process variable registration unit 3 and appropriately normalizes it using the average or variance of each variable. In addition, the outlier is removed as necessary. Then, the current Q statistic and the T ² statistic are monitored by substituting into the Q statistic defined by the formulas (3) and (4) and the T ² statistic X (t). Since this statistic changes from time to time, it may be monitored in the form of a time series graph (trend graph). Then, when the current Q statistic or T ² statistic exceeds the threshold value defined by the equations (5) and (6), it is determined that a state change has occurred in the process. This is an example of the operation of the abnormality detection unit 72.

Next, in the abnormality factor variable separation unit 73, when the abnormality detection unit 72 detects a process abnormality, the abnormality detection data at that time, that is, the Q statistics of the equations (3) and (4) at that time And the T ² statistic are input, and the contribution amounts of the equations (7) and (8) are calculated. Then, for each process variable, the order of which variable contribution amount is large is ranked. This is an example of the operation of the abnormal factor variable separation unit 73.

  The operation of the abnormality detection unit 72 and the operation of the abnormality factor variable separation unit 73 are the same as the operation of MSPC using ordinary PCA.

  Next, the abnormality countermeasure support information providing unit 74 uses the process variable data ranked by the abnormality factor variable separating unit 73 and the abnormality countermeasure support message template defined by the abnormality countermeasure support information defining unit 56. Provide countermeasure support information in case of abnormality.

  In order to show a specific implementation image, the operation of generating a simplified message will be described with reference to FIG. In FIG. 10, a simplified version of the process variable correlation diagram of FIG. 6 is used to simplify the description.

  First, it is assumed that the discharge TP, the air volume, and the dissolved oxygen concentration DO are extracted as the upper abnormal factor candidates FX. The discharge TP is classified as an output variable FY, the air volume is classified as an input variable FU, and the dissolved oxygen concentration DO is classified as an intermediate variable FZ. Then, as shown in FIG. 11, a countermeasure support message is generated using the countermeasure support message template at the time of abnormality defined by the countermeasure countermeasure definition information 56 at the time of abnormality.

  In the present embodiment, the simplest example in which the abnormal factor candidate variable FX is classified into the output variable FY, the intermediate variable FZ, and the input variable FU one by one and they are all related is shown. In addition, a support message is automatically provided to a user such as an operator through the user interface unit 9 in accordance with a countermeasure support message template at the time of abnormality defined by the countermeasure support information definition unit 56 at the time of abnormality. The action of providing such a message is the action of the embodiment of the abnormality countermeasure support information providing unit 74. Thereby, the operation of the abnormality monitoring / diagnosis / countermeasure support unit 7 is completed.

  Next, the user who is provided with the abnormality countermeasure support information performs some operation based on the support information, but it is more preferable to save a countermeasure record such as an operation record. This is the function of the abnormality countermeasure record holding unit 8. For example, when the air volume is adjusted in accordance with the message, a record that the air volume is adjusted is automatically saved. Further, there is a column on the user interface unit 9 that can be input by the operator, how to adjust the air volume, and as a result, how much improvement in the discharge TP is recognized, and the adverse effect of adjusting the air volume. (Records of side effects) etc. should be recorded. In this abnormality countermeasure record holding unit 8, it is possible to keep a record of which operation amount has been adjusted, a record of which performance measure is the countermeasure, and a record of which management item is the countermeasure. It is more preferable to have a function that allows a search to be performed by a performance index or a management item. In addition, it is better to have a function that allows the operator's records to be searched with keywords. This is an operation example of the embodiment of the abnormality countermeasure record holding unit 8.

  If this function is provided, the abnormality countermeasure support information providing unit 74 can automatically search for this function and display a past record list of a certain operation amount, performance index, or management index. Good. For example, it is preferable that a list of past abnormality countermeasures at the time of discharge phosphorus concentration, a past abnormality countermeasure list at the time of abnormal air flow, and a list of past abnormality countermeasures at the time of dissolved oxygen concentration DO abnormality can be presented. This is an operation example corresponding to claim 8.

Thus, the operation of abnormality monitoring, diagnosis, and countermeasure support is completed.

  As described above, according to the present embodiment, the process monitoring / diagnosis / support apparatus includes the abnormality diagnosis model construction unit 5 and the abnormality monitoring / diagnosis / countermeasure support unit 7, so that the abnormality is not dependent on the target process. Time support information can be provided.

  In addition, the conventional monitoring / diagnostic methods can only detect abnormal signs and separate the cause variables, or have required a lot of engineering to provide countermeasure support information. If the method of the form is used, the countermeasure support information at the time of abnormality can be provided almost mechanically in addition to abnormality detection and factor variable separation with almost no engineering.

  In other words, according to the present embodiment, process variables to be input to the diagnostic system are defined and collected by SCADA or the like with little intervention other than classifying the process variables into input variables, output variables, and intermediate variables. The plant operator is not dependent on the target process, but only by passing the data processing mechanically (automatically without engineering intervention) from the monitoring data of the plant being used. It is possible to provide countermeasure support information to users such as a plant manager.

  In addition, the names of n process variables are displayed, and during this display, the process variable classification user interface unit 4 selects an input variable, an output variable, and an intermediate variable in the process variable classification unit 51 according to a user operation. , The user can easily classify the process variables on the GUI.

  Further, the data definition unit 54 for anomaly detection includes principal component analysis (PCA), latent variable projection method (PLS), principal component regression (PCR), canonical correlation analysis (CVA), Mahalanobis distance, and robust principal component analysis ( Robust PCA), kernel principal component analysis (Kernel PCA), discriminant analysis, fuzzy C-means clustering, K-means clustering, support vector machine (SVM), relevance vector machine (RVM), and subspace method When defining an expression for generating anomaly detection data using more than one method, the existing multivariate effect of providing countermeasure support information during non-stationary conditions in a form that does not depend on the target process It can be provided systematically with the aid of analysis and machine learning techniques. In addition, based on MSPC, which is a statistical FDI technique that does not require knowledge related to the target process, it shows what measures should be taken for operators and managers when the process is unsteady or abnormal Specific countermeasure support information can be provided.

  Also, the abnormality detection data contribution amount definition unit 55 calculates the contribution amount of the target process variable k among the n process variables (where k = 1, 2,..., N) as [a] abnormality detection. It is defined by projecting from the abnormality detection data defined by the data definition unit 54 to the space spanned by the process variable k, or [b] decomposing the abnormality detection data into the sum of n components, When the variable k is defined as the k-th component to be affected or [c] is defined using the independent component analysis, the countermeasure support information in the non-stationary state is provided without depending on the target process. The above-described effect of the effect can be further provided systematically.

  In addition, the process variable correlation definition unit 53 defines the relationship between process variables when defining correlation information based on a correlation matrix or principal component load matrix of time-series data over a predetermined period of n process variables. Presence / absence or strength can be easily defined mechanically and systematically from the data. In addition, since many of the methods such as multivariate analysis in the anomaly detection data definition unit 54 perform an operation of creating a variance-covariance matrix and a principal component load matrix at the time of anomaly detection data creation, You can mechanically define the relationship between process variables with a simple modification.

  Further, the process variable correlation definition unit 53 has a table of n rows and n columns indicating the presence or absence of the correlation between n process variables or the strength as correlation information, and the process variable correlation definition interface unit 6 performs a user operation. Accordingly, when the table is displayed and the contents of the table are corrected, the above-described effects of the process monitoring / diagnosis / support apparatus constructed by data driven cannot be satisfied by users such as plant managers and operators. In some cases or when the user wants to make corrections, the process variable correlation definition user interface unit 6 can easily make corrections.

  Further, when the second to sixth rules are defined in the abnormality countermeasure support information definition unit 56, as a more specific form of the above-described effect, the user confirms which operation amount at the time of abnormality and adjusts as necessary. Support information on what to do, and what kind of process monitoring data is likely to be affected by this amount of operation (eg, performance indicators such as effluent quality may change in the future) Can be provided to the user.

  Further, when the abnormality countermeasure record holding unit 8 is provided, when the support information is provided, the plant operator / manager can refer to what countermeasure has been taken in the past similar cases. This makes it easy to take measures.

<Configuration of Other Embodiments>
12 to 14 respectively implement the monitoring system for the sewage advanced treatment process 1 of FIG. 1 described in the first embodiment through a communication line such as a public line such as the Internet or a telephone or a dedicated line. 1 shows an example of a system configuration. The communication line is wired or wireless, but it is assumed that the communication line is mainly realized by a wireless line. The components in the configuration of the embodiment in FIGS. 12 to 14 are substantially the same as the components in the configuration in FIG. 1 except that the functions are divided through a communication line. In FIG. 14, the remote monitoring operation terminal 10 is newly provided.

  In the configuration shown in FIG. 12, the collection of data measured in the sewage altitude treatment process 1 and the operation of each manipulated variable by the various actuators 111 to 115 in the sewage altitude treatment process 1 are performed at the treatment plant. Is implemented in the center apparatus 11 that exists in a physically separated place through a communication line.

  That is, the configuration shown in FIG. 12 further includes a center device 11 connected to the process measurement data collection / storage unit 2 through a wired or wireless communication line. Each part 3, 4, 5, 6, 7, 9 is mounted on the center device 11. The center apparatus 11 collects time series data of n process variables stored in the process measurement data collection / storage unit 2 through a communication line, presents them to the user, and operates the input variable U according to the user's operation. In the case, the operation signal of the input variable U is transmitted to the sewage altitude treatment process 1 through the communication line.

  The configuration shown in FIG. 13 differs from FIG. 12 in that the remote monitoring operation terminal receives the countermeasure support information at the time of abnormality through the communication line and transmits the result determined based on this information to the sewage altitude treatment process 1 again through the communication line. 10. A plurality of remote monitoring operation terminals 10 may exist, and a plurality of plant managers and operators who perform operation management of the sewage advanced treatment process 1 can have this terminal.

  That is, the configuration shown in FIG. 13 further includes a remote monitoring operation terminal 10 connected to the abnormality countermeasure support information providing unit 74 and the sewage altitude processing process 1 through a wired or wireless communication line. When the monitoring operation terminal 10 presents the countermeasure support information provided from the abnormality countermeasure support information providing unit 74 through the communication line to the user and operates the input variable U according to the user's operation, the monitoring operation terminal 10 An operation signal is transmitted to the sewage advanced treatment process 1 through a communication line.

  The configuration shown in FIG. 14 is a configuration developed by combining FIG. 12 and FIG. 13. In the advanced sewage treatment process 1, data collection and operation amount manipulation are performed, and diagnosis is performed at the center through a communication line. Is done. In this center, diagnosis for a plurality of processing processes is performed. Furthermore, it has a remote monitoring operation terminal 10, which is diagnosed at the center, transmits countermeasure support information to this remote monitoring operation terminal 10 through the communication line, and again from the remote monitoring operation terminal 10 through the communication line, An operation command for the processing process can be performed.

  That is, in the configuration shown in FIG. 14, when there are w sewage altitude treatment processes 1 (where w = 1, 2,... W), w process measurement data collection / storage units 2 are provided, and the w processes The measurement data collection / storage unit 2 is individually arranged in the sewage advanced treatment process 1 at w locations. In addition, the center apparatus 11 connected to each process measurement data collection / storage unit 2 through a wired or wireless communication line, and the abnormality countermeasure support information providing unit 74 and each sewage advanced treatment process 1 individually through a wired or wireless communication line And w or more remote monitoring operation terminals 10 connected to the network. The center apparatus 11 collects time-series data of n process variables stored in each process measurement data collection / storage unit 2 through a communication line and presents them to a first user (a user of the center apparatus 11). When operating the input variable U according to the user's operation, the operation signal of the input variable U is transmitted to the sewage altitude treatment process 1 through the communication line. Each remote monitoring operation terminal 11 presents countermeasure support information provided from the abnormality countermeasure support information providing unit 74 through the communication line to the second user (user of the remote monitoring operation terminal 11), and responds to the operation of the second user. When operating the input variable U, the operation signal of the input variable U is transmitted to the sewage altitude treatment process 1 through the communication line.

  Next, the operation of the other embodiment configured as described above will be described.

  In addition, since the basic effect | action of other embodiment is the same as the effect | action of 1st Embodiment, it mainly describes a different part.

  First, the operation of the embodiment having the configuration shown in FIG. 12 will be described. In the center apparatus 11, the process variable registration unit 3 registers in advance process variables necessary for constructing the abnormality diagnosis model, and sends this registration information to the processing place. The processing site holds information on the registered process variables so that it can be referred to as appropriate. In addition, the past (offline) data extraction unit 52 of the abnormality diagnosis model construction unit 5 and the current (online) data extraction unit 71 of the abnormality monitoring / diagnosis / measure support unit 7 collect process measurement data on the sewage advanced treatment process 1 side. It is held in the form associated with the storage unit 2.

  Next, an abnormality diagnosis model is constructed. Time series data of registered variables extracted by the past (offline) data extraction unit 52 over a predetermined period is transmitted to the center apparatus 11 through the communication line, and the center apparatus 11 performs the first implementation. An abnormality diagnosis model is constructed according to the procedure shown in the operation of the form.

  Next, the abnormality monitoring / diagnosis / countermeasure support unit 7 performs diagnosis at the time of diagnosis (hereinafter referred to as “current”). Here, first, the current (online) data extraction unit 71 extracts the data of the current registration variable in the processing plant, and sends it to the center apparatus 11 through the communication line. In the center apparatus 11, abnormality detection / factor separation / support information generation is performed according to the procedure shown in the operation of the first embodiment, which is presented to a user such as an operator through the user interface unit 9.

  Next, based on the countermeasure support information, the operator confirms the performance index and management value of the target process, and if there is a need to change the operation amount, determines an operation amount change command and communicates this information. Report to advanced sewage treatment process 1 through the line

  In the sewage altitude treatment process 1, the manipulated variable is changed automatically or manually based on the received command information. When the treatment plant is unmanned, the amount of operation is automatically changed. When the treatment plant is manned, the operator of the treatment plant confirms the command information and changes the operation amount manually or automatically.

  In addition, although the center apparatus assumes the computer arrange | positioned in the monitoring room installed in the place physically separated from the processing place, as long as it has the function implemented in the center apparatus 11, For example, the units 3 to 7 and 9 may be mounted on a mobile terminal such as a mobile phone or a mobile PC.

  This is the operation of the embodiment having the configuration shown in FIG.

  Next, the operation of the embodiment having the configuration shown in FIG. 13 will be described.

  First, in the treatment plant with the sewage advanced treatment process 1, an abnormality diagnosis model is constructed according to the procedure shown in the operation of the first embodiment. Subsequently, in the treatment plant with the same sewage advanced treatment process 1, the first Abnormality detection, factor separation, and generation of support information are performed according to the procedure shown in the operation of the embodiment. Up to this point, the operation is exactly the same as in the first embodiment.

  Next, only the generated countermeasure support information is transmitted to the plant operator or manager having the remote monitoring operation terminal 10 through the communication line. At this time, a plurality of operators and managers may have the remote monitoring operation terminal 10, and in this case, countermeasure support information is provided by a broadcast method.

  Next, the operator or plant manager who has received this information determines whether or not the manipulated variable needs to be changed, determines the manipulated variable change command if necessary, and processes this information through the communication line to treat the sewage altitude. Tell process one. At this time, when a plurality of persons have the remote monitoring operation terminal 10, it is desirable to have a priority order with respect to the operation amount command indicating which terminal gives priority to the command. In this method, since only countermeasure support information is received, the performance index and management value of the target process cannot be confirmed, but the performance index that the operator wants to confirm through the remote monitoring operation terminal 10 only when the abnormality countermeasure support information is received. It is desirable to have a function of making a request for transmission of management value data to the processing site.

  Finally, in the sewage altitude treatment process 1, the manipulated variable is changed automatically or manually based on the received command information. When the treatment plant is unmanned, the amount of operation is automatically changed. When the treatment plant is manned, the operator of the treatment plant confirms the command information and changes the operation amount manually or automatically.

  In this embodiment, the remote monitoring operation terminal 10 is mainly assumed to be a mobile terminal such as a mobile phone or a mobile PC, but a center such as a monitoring room existing in a physically fixed place having the above-described mechanism. The essential difference from the embodiment of FIG. 12 is how to divide the function sharing.

  This is the operation of the embodiment having the configuration shown in FIG.

  Next, the operation of the embodiment having the configuration shown in FIG. 14 will be described.

  In the center apparatus 11, the process variable registration unit 3 registers in advance process variables necessary for constructing the abnormality diagnosis model, and sends this registration information to the processing place. The difference from the embodiment of FIG. 12 is that there are a plurality of processing sites, and the center apparatus 11 remotely monitors the plurality of processing sites collectively. Accordingly, the process variables are registered for the number of processing sites, but this may be defined for each processing site or a common process variable may be registered. Each processing place holds information on the registered process variables so that it can be referred to as appropriate. In addition, the past (offline) data extraction unit 52 of the abnormality diagnosis model construction unit 5 and the current (online) data extraction unit 71 of the abnormality monitoring / diagnosis / measure support unit 7 collect process measurement data on the sewage advanced treatment process 1 side. It is held in the form associated with the storage unit 2.

  Next, an abnormality diagnosis model is constructed. Time series data of registered variables over a predetermined period extracted by the past (offline) data extraction unit 52 is transmitted from each processing site to the center apparatus 11 through the communication line. In accordance with the procedure shown in the operation of the first embodiment, an abnormality diagnosis model for each processing site is constructed.

  Next, the abnormality monitoring / diagnosis / countermeasure support unit 7 performs diagnosis at the time of diagnosis (hereinafter referred to as “current”). Here, first, the current (online) data extraction unit 71 extracts the data of the current registered variable in each processing place, and this is sent to the center apparatus 11 through the communication line. In the center apparatus 11, the abnormality detection / factor separation / support information generation is performed in accordance with the procedure shown in the operation of the first embodiment. At this time, the abnormality diagnosis model to be referred to needs to be switched for each processing place, but the operation of detection / diagnosis / countermeasure support can be implemented as one algorithm other than the switching of the diagnosis model.

  When the countermeasure support information is generated, the generated information is presented to a user such as an operator through the user interface unit 9. At this time, an operator or manager in a place other than the monitoring room where the center device 11 is located may have the remote monitoring operation terminal 10, and in this case, the countermeasure support information generated at the same time is remotely transmitted through the communication line. It is also transmitted to the plant operator or manager who owns the monitoring operation terminal 10. At this time, a plurality of operators and managers may have the remote monitoring operation terminal 10, and in this case, countermeasure support information is provided by a broadcast method.

  Next, based on the countermeasure support information, the operator confirms the performance index and management value of the target process, and if there is a need to change the operation amount, determines an operation amount change command and communicates this information. Report to advanced sewage treatment process 1 through the line When there is an operator or manager who owns the remote monitoring operation terminal 10, which command is given priority to the center or the remote terminal, or when there are a plurality of remote terminal owners, which command is given priority It is desirable to have a priority order for the manipulated variable command.

  Finally, in the sewage altitude treatment process 1, the manipulated variable is changed automatically or manually based on the received command information. When the treatment plant is unmanned, the amount of operation is automatically changed. When the treatment plant is manned, the operator of the treatment plant confirms the command information and changes the operation amount manually or automatically.

  The above is the operation of the embodiment having the configuration shown in FIG.

  The effects of the embodiment shown in FIGS. 12 to 14 are as follows.

  According to the configuration shown in FIG. 12, even when the sewage treatment plant and the plant operator or plant manager that operates and manages the sewage treatment plant are physically separated, the same effect as that of the first embodiment can be obtained. Can be obtained.

  According to the configuration shown in FIG. 13, the mobile terminal is designed to reduce the calculation load on the center side in the mechanism of FIG. 12, and in addition to the effects of the first embodiment, it is difficult to apply a large calculation load. However, it is possible to remotely implement information and countermeasures for abnormalities that are most necessary for process monitoring. In addition, because the communication line is used only when there is an abnormality, the load on the communication line can be reduced at the same time, contributing to avoiding traffic congestion on the communication line, and important information even when the traffic is congested ( It is easy to ensure communication of information at the time of abnormality.

  Supplementally, according to the configuration shown in FIG. 13, even when the plant operator or the plant manager is physically separated from the target plant and traffic congestion of the communication line is expected, the target It is possible to detect and deal with important anomalies that occur in the process (however, at the cost of giving up monitoring time series data for all the process variables being measured).

  According to the configuration shown in FIG. 14, the configuration of FIGS. 12 and 13 is further divided into functions, and it is intended to collectively monitor a plurality of treatment plants, but such a configuration is adopted. By doing so, the effect of 1st Embodiment can be provided as a remote service (cloud service).

  That is, according to the configuration shown in FIG. 14, monitoring / diagnosis / support of a plurality of plants can be carried out in the center apparatus 11 arranged in one or a small number of monitoring rooms. The effect of the first embodiment can be realized. In addition, even when a plant operator or a plant manager is physically away from the target plant, plant monitoring, diagnosis, and countermeasure support can be performed.

  According to at least one embodiment described above, the process monitoring / diagnosis / support device includes the abnormality diagnosis model construction unit 5 and the abnormality monitoring / diagnosis / countermeasure support unit 7 in a form that does not depend on the target process, It is possible to provide countermeasure support information in case of abnormality.

  In addition, although some embodiment of this invention was described, these embodiment is shown as an example and is not intending limiting the range of invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 1 ... Advanced sewage treatment process, 101 ... First sedimentation tank, 102 ... Anaerobic tank, 103 ... Anoxic tank, 104 ... Aerobic tank, 105 ... Final sedimentation tank, 2 ... Process measurement data collection / storage unit, 3 ... Process variable Registration unit, 4 ... process variable classification user interface unit, 5 ... abnormality diagnosis model construction unit, 51 ... process variable classification unit, 52 ... past (offline) data extraction unit, 53 ... process variable correlation definition unit, 54 ... for abnormality detection Data definition unit, 55 ... Data contribution definition unit for anomaly detection, 56 ... Error countermeasure support information definition unit, 6 ... Process variable correlation definition interface unit, 7 ... Anomaly monitoring / diagnosis / countermeasure support unit, 71 ... Present (online) ) Data extraction unit, 72 ... abnormality detection unit, 73 ... abnormality factor variable separation unit, 74 ... abnormality countermeasure support information provision unit, 8 ... abnormality countermeasure record holding unit, 9 ... user-in Over the face portion, 10 ... remote monitoring control terminal, 111 to 115 ... actuator and the operation amount sensors, 121-1221 ... process sensor.

Claims (12)

Data collection and storage means for collecting and storing time-series data of n process variables indicating measurement results from n (where n ≧ 2) sensors that measure state quantities or manipulated variables of the target process;
The n process variables are represented by p (where 1 ≦ p <n) output variables Y representing the performance indicators and L (where 1 ≦ L <n) representing the manipulated variables. ) Input variable U and process variable classification means for storing classification information indicating the result of classification into m (where 0 ≦ m <n) intermediate variables Z representing the state quantities and representing management / monitoring values When,
A process variable correlation defining means for defining correlation information indicating correlation between the n process variables including the classified result from the time series data of the classification information and the n process variables over a predetermined period; ,
From the time-series data of the n process variables over a predetermined period, q (where 1 ≦ q << n) abnormality detection data that is less than n is generated, and the abnormality detection data Anomaly detection data defining means for defining a judgment criterion for judging the presence or absence of anomaly or anomaly intensity based on;
Anomaly detection data contribution amount defining means for defining contribution amounts of time series data of the n process variables to the anomaly detection data;
Based on the correlation information, the first rule for extracting a predetermined higher-order abnormality factor candidate variable among the abnormality factor candidate variables that are process variables that are abnormality factor candidates at the time of abnormality, and the extracted abnormality factor candidate variable based on the correlation information A second rule that classifies the output variable Y, the input variable U, and the intermediate variable Z; and a third rule that provides countermeasure support information including a message that supports countermeasures in case of abnormality according to the classified result. An error countermeasure support information defining means to be defined;
The abnormality detection data corresponding to the time-series data at the predetermined time is generated according to the stored time-series data of the n process variables at the predetermined time and the definition of the abnormality detection data definition means, and the abnormality detection Abnormality detecting means for diagnosing the presence or absence of abnormality of the data at the predetermined time based on the data for use,
As a result of the diagnosis, if there is an abnormality, the contribution amount of the time series data at the predetermined time is calculated based on the definition of the contribution amount, and the predetermined time is calculated based on the calculated contribution amount and the first rule. Anomaly factor variable separating means for separating and extracting anomaly factor candidate variables from time series data;
Anomaly countermeasure support information providing means for providing the countermeasure support information based on the extracted and extracted abnormality factor candidate variable, the correlation information, and the second rule and the third rule;
A process monitoring / diagnosis / support device characterized by comprising: The process monitoring / diagnosis / support apparatus according to claim 1,
The process variable classification user interface means for displaying the names of the n process variables and selecting an input variable, an output variable, and an intermediate variable in the process variable classification means according to a user operation during the display. Process monitoring / diagnosis / support device. The process monitoring / diagnosis / support apparatus according to claim 1,
The anomaly detection data definition means includes principal component analysis (PCA), latent variable projection (PLS), principal component regression (PCR), canonical correlation analysis (CVA), Mahalanobis distance, and robust principal component analysis (Robust PCA). At least one of kernel principal component analysis (Kernel PCA), discriminant analysis, fuzzy C-means clustering, K-means clustering, support vector machine (SVM), relevance vector machine (RVM), and subspace method A process monitoring / diagnosis / support apparatus characterized by defining an expression for generating the abnormality detection data using a method. In the process monitoring / diagnosis / support apparatus according to claim 3,
The abnormality detection data contribution amount defining means determines the contribution amount of the process variable k of interest (where k = 1, 2,..., N) among the n process variables as [a] the abnormality detection. It is defined by projecting from the anomaly detection data defined by the data definition means to the space spanned by the process variable k, or [b] decomposing the anomaly detection data into a sum of n components, and the process variable k A process monitoring / diagnosis / support apparatus characterized in that it is defined as a k-th component that only affects or [c] independent component analysis. The process monitoring / diagnosis / support apparatus according to claim 1,
The process variable correlation defining means defines the correlation information based on a correlation matrix or principal component load matrix of time-series data over a predetermined period of the n process variables. Support device. The process monitoring / diagnosis / support apparatus according to claim 1,
The process variable correlation defining means includes a table of n rows and n columns representing the presence or absence or strength of correlation between the n process variables as the correlation information,
A process monitoring / diagnosis / support apparatus characterized by further comprising a process variable correlation definition interface means (6) for displaying the table according to the user's operation and correcting the contents of the table. The process monitoring / diagnosis / support apparatus according to claim 1,
In the second rule, the extracted abnormal factor candidate variables are classified into the input factor variable FU belonging to the input variable U, the output factor variable FY belonging to the output variable Y, and the intermediate variable Z based on the classification information. Including rules for classifying the intermediate factor variable FZ to which it belongs,
The third rule is:
When the abnormal cause candidate variable includes the output factor variable FY, it is notified that the output factor variable FY has deteriorated, and the correlation information is defined as correlated with the output factor variable FY. A fourth rule A for providing countermeasure support information including a message prompting the user to check and review the value of the input variable U;
If the abnormal factor candidate variable does not include the output factor variable FY but includes the input factor variable FU, the input factor variable FU is in an abnormal state deviating from the normal time, and is represented by the input factor variable FU. And providing countermeasure support information including a message notifying that the output variable Y defined as correlated with the input factor variable FU may be deteriorated in the future by the correlation information. The fifth rule B;
When the abnormal factor variable is only the intermediate factor variable FZ, it is notified that the intermediate factor variable FZ has deteriorated, and the correlation information is defined as correlated with the intermediate factor variable FZ. A sixth rule C for providing countermeasure support information including a message notifying that the output variable Y may deteriorate in the future;
A process monitoring / diagnosis / support device characterized by comprising: The process monitoring / diagnosis / support apparatus according to claim 1,
After presenting the countermeasure support information provided by the abnormality countermeasure support information providing means, further comprising an abnormality countermeasure record holding means for recording the operation performed by the user,
The abnormality countermeasure record holding means is:
When the countermeasure support information is provided, based on the time series data and the classification information, the value of the input variable U, the value of the output variable Y and the value of the intermediate variable Z before the current time, the countermeasure support information, Means for presenting,
After presenting the countermeasure support information, the operation record of the input variable U as the operation performed by the user, the value of the output variable Y, and the value of the intermediate variable Z are continuously recorded. Monitoring / diagnosis / support equipment. The process monitoring / diagnosis / support apparatus according to claim 1,
A center device connected to the data collection and storage means through a wired or wireless communication line;
The process variable classification means, the process variable correlation definition means, the abnormality detection data definition means, the abnormality detection data definition means, the abnormality detection data contribution amount definition means, the abnormality detection means, and the abnormality factor variable separation means And the abnormality countermeasure support information providing means is mounted on the center device,
The center apparatus collects time series data of n process variables stored in the data collection and storage means through the communication line, presents them to the user, and operates the input variable U according to the user's operation. In this case, the process monitoring / diagnosis / support apparatus is characterized by transmitting an operation signal of the input variable U to the target process through the communication line. The process monitoring / diagnosis / support apparatus according to claim 1,
A monitoring operation terminal connected to the abnormality countermeasure support information providing means and the target process through a wired or wireless communication line;
The monitoring operation terminal presents countermeasure support information provided from the abnormality countermeasure support information providing means through the communication line to a user, and operates the input variable U in response to the user's operation. A process monitoring / diagnosis / support apparatus, wherein an operation signal of a variable U is transmitted to the target process through the communication line. The process monitoring / diagnosis / support apparatus according to claim 1,
When there are w target processes (where w = 1, 2,... W), w data collection / save means are provided, and the w data collection / save means are individually arranged in the w target processes. And
A center device connected to each data collection and storage means through a wired or wireless communication line;
The abnormality countermeasure support information providing means and w or more monitoring operation terminals individually connected to each target process through a wired or wireless communication line,
The center device collects time series data of n process variables stored in the data collection and storage means through the communication line, presents the data to the first user, and inputs the input according to the operation of the first user. When operating the variable U, the operation signal of the input variable U is transmitted to the target process through the communication line,
Each monitoring operation terminal presents countermeasure support information provided from the abnormality countermeasure support information providing means through the communication line to a second user, and operates the input variable U according to the operation of the second user. Includes a process monitoring / diagnosis / support apparatus that transmits an operation signal of the input variable U to the target process through the communication line. A program used in a process monitoring / diagnosis / support device with a memory,
The process monitoring / diagnosis / support device;
Data collection and storage means for collecting time-series data of n process variables indicating measurement results from n (where n ≧ 2) sensors that measure state quantities or manipulated variables of the target process and storing them in the memory;
The n process variables are represented by p (where 1 ≦ p <n) output variables Y representing the performance indicators and L (where 1 ≦ L <n) representing the manipulated variables. ) Input variable U and process variable classification means for storing classification information indicating the result of classification into m (where 0 ≦ m <n) intermediate variables Z representing the state quantities and representing management / monitoring values ,
A process variable correlation defining means for defining correlation information indicating correlation between the n process variables including the classified result from time-series data over a predetermined period of the classification information and the n process variables;
From the time-series data of the n process variables over a predetermined period, q (where 1 ≦ q << n) abnormality detection data that is less than n is generated, and the abnormality detection data Anomaly detection data defining means for defining a judgment criterion for judging the presence or absence of anomaly or anomaly intensity based on
Anomaly detection data contribution amount defining means for defining a contribution amount of time series data of the n process variables to the anomaly detection data;
Based on the correlation information, the first rule for extracting a predetermined higher-order abnormality factor candidate variable among the abnormality factor candidate variables that are process variables that are abnormality factor candidates at the time of abnormality, and the extracted abnormality factor candidate variable based on the correlation information A second rule that classifies the output variable Y, the input variable U, and the intermediate variable Z; and a third rule that provides countermeasure support information including a message that supports countermeasures in case of abnormality according to the classified result. An emergency countermeasure support information definition means to be defined,
The abnormality detection data corresponding to the time-series data at the predetermined time is generated according to the stored time-series data of the n process variables at the predetermined time and the definition of the abnormality detection data definition means, and the abnormality detection Abnormality detecting means for diagnosing the presence or absence of abnormality of the data at the predetermined time based on the data for use,
As a result of the diagnosis, if there is an abnormality, the contribution amount of the time series data at the predetermined time is calculated based on the definition of the contribution amount, and the predetermined time is calculated based on the calculated contribution amount and the first rule. Anomalous factor variable separating means for separating and extracting abnormal factor candidate variables from time series data,
Anomaly countermeasure support information providing means for providing the countermeasure support information based on the separated and extracted abnormality factor candidate variable, the correlation information, and the second rule and the third rule;
Program to function as.

Images